Do tech companies need to seek permission before they train an AI model with copyrighted data? Yes, ruled one federal judge in February, concluding a copyright infringement lawsuit that may define the many legal battles to come. Information and software company Thomson-Reuters filed the lawsuit in 2020, alleging that the legal startup Ross Intelligence violated copyright when it trained its AI model with editorial content from Westlaw, a Thomson-Reuters-owned legal search engine.
According to the Verge, Ross's defense — that permission was not required because their activities should be considered fair use — fell flat with Judge Stephanos Bibas. In his occasionally scathing opinion, he agreed with Thomson-Reuters' claim that Ross used copyrighted data to built a direct competitor to the Westlaw service, and noted that "none of Ross's possible defenses hold water."
The ruling could potentially influence the outcomes of dozens of AI-related copyright infringement cases currently winding their way through courthouses, including a recently filed lawsuit brought by five major media companies against the startup Cohere. According to Betakit, the lawsuit alleges "massive, systemic copyright infringement" and accuses Cohere of using their copyrighted content to train its AI models and generate falsely attributed fake articles.
It's just one battle in what is likely to be a long and arduous war. But so far, copyright holders seem to hold a slight advantage.
